Table 2.

Risk of bias of reviewed studies.

Articles/Bias	Selection bias		Performance bias		Detection bias		Attrition bias		Reporting bias
	Risk	Text where it is located	Risk	Text where it is located	Risk	Text where it is located	Risk	Text where it is located	Risk	Text where it is located
Daley et al. (2018)	Low	“At the first session, participants were randomly assigned (based on a computer-generated code)…”	High	“It was not feasible to mask participants or researchers to group allocation.”	Unc.	“…higher self-reports of activity in the physical activity group compared with the control group may be biased by knowledge of allocation.”	Low	“When using an intention to treat approach it is acceptable to exclude patients' data, without risking bias…”	Low	The study protocol is available and all of the study's pre-specified (primary and secondary) outcomes that are of interest in the review have been reported.
Vargas-Terrones et al. (2019)	Low	“A simple randomization was performed with the Epidat V.3.1 program to allocate the participants into two groups in order of entry: intervention group (IG) and control group (CG).”	High	The nature of the intervention prevented the study from blinding participants.	Low	“Participants were not involved in the design, recruitment, and conduct of the study.”	Low	“An analysis by intention-to-treat was also performed using two different methods.”	Low	The study protocol is available and all of the study's pre-specified (primary and secondary) outcomes that are of interest in the review have been reported.
Haakstad et al. (2016)	Low	“Allocations were sealed in opaque numbered envelopes following a simple computer-based randomization program.”	High	The nature of the intervention prevented the study from blinding participants.	Low	“In order to treat the two groups identically apart from for the experimental intervention, the controls underwent all tests and completed the same interview as the exercise group, also with respect to assessment of total physical activity level and exercise habits.”	Low	“The principal analysis was done on an intention to treat basis…”	Low	The study protocol is available and all of the study's pre-specified (primary and secondary) outcomes that are of interest in the review have been reported.
Perales et al. (2016)	Low	“Thereafter, they were randomly assigned to a standard care (control, initial n = 121) or intervention group (exercise, n = 120) using a computer-generated list of random Numbers.”	High	“The study participants and the qualified fitness instructors who supervised the exercise sessions were not blinded to the group allocation.”	Low	“The researchers responsible for assessing eligibility, baseline measures, or outcome assessment were blinded to the group allocation.	Low	“All the analyses were performed using the Statistical Package for Social Sciences program version 22.0, and were adhered to the intention-to-treat principle…”	Low	The study protocol is available and all of the study's pre-specified (primary and secondary) outcomes that are of interest in the review have been reported.
Uebelacker et al. (2016)	Low	“Once we received clearance, we re-contacted the participant and randomized her to the prenatal yoga program (PYP) or a perinatal health education control condition.”	High	“Because this is a study of behavioral interventions, participants could not be blind to which intervention they received.”	Low	“Study groups did not differ on any variables.”	Low	No missing data.	High	Cannot locate protocol.
Davis et al. (2015)	Low	“Participants in both conditions completed a clinical interview and baseline self-report questionnaires prior to randomization.”	Unc.	“Interrater reliability for the Yoga Adherence Scale was 98% for the four classes that both research assistants evaluated.”	Unc.	There is no evidence about blinding of outcomes, but everything suggests that it is done.	Low	No missing data.	High	Cannot locate protocol.
Perales et al. (2015)	Low	“A computer-generated list of random numbers was used to allocate the participants into the groups.”	Unc.	“The randomization blinded process (sequence generation, allocation concealment, and implementation) was performed by three different authors.”	Unc.	There is no evidence about blinding of outcomes, but everything suggests that it is done.	Low	No missing data.	Low	The study protocol is available and all of the study's pre-specified (primary and secondary) outcomes that are of interest in the review have been reported.
Ussher et al. (2015)	Low	“An independent statistician generated a randomization list using Stata, with random permuted blocks of random size stratified by recruitment cente, in a 1:1 ratio.”	High	“It was not feasible to mask participants or researchers to group allocation.”	Unc.	There is no evidence about blinding of outcomes, but everything suggests that it is done.	Low	“Analysis was on an intention to treat basis…”	Low	The study protocol is available and all of the study's pre-specified (primary and secondary) outcomes that are of interest in the review have been reported.
Perales et al. (2014)	Low	“For allocation of the participants, a computer-generated list of random numbers was used.”	High	“…due type of intervention blinding of participants was not possible.”	Low	“Randomization process (sequence generation, allocation concealment, and implementation) was made for three different authors in order to facilitate blinding of process and outcomes assessment.”	Low	No missing data.	Low	The study protocol is available and all of the study's pre-specified (primary and secondary) outcomes that are of interest in the review have been reported.
Field et al. (2013a)	Low	“…the depressed pregnant women were randomly assigned to a yoga or a social support group based on a random numbers table.”	Unc.	“…by trained researchers who were blinded to the group assignment and the study hypotheses.	Low	“The groups did not differ on demographic variables and baseline measures.	Low	No missing data.	High	Cannot locate protocol.
Field et al. (2013b)	Low	“The participants were clinically depressed pregnant women who were randomly assigned to either a tai chi/yoga treatment or a control group.”	Unc.	The nature of the intervention prevented the study from blinding participants.	Unc.	There is no evidence about blinding of outcomes, but everything suggests that it is done.	Low	No missing data.	High	Cannot locate protocol.
Satyapriya et al. (2013)	Low	“…the subjects were allocated to two groups (yoga and control) using a computer generated random number…”	Unc.	As this was an interventional study, the participants or the trainer could not be blinded. Attempts were made to mask wherever feasible to reduce the bias.	Low	“The team who did the assessments was not involved in administering the intervention. The statistician who did the randomization and analysis was blind to the source of the data.”	Low	No missing data.	High	Cannot locate protocol.
Field et al. (2012)	Low	“The women were then randomly assigned to a yoga, massage therapy or standard prenatal care control group.”	Unc.	“All assessments were conducted by the trained research associates who were blind to the study' hypotheses and to the group assignment.”	Low	“In addition, the yoga and massage therapy groups did not differ on neonatal outcomes including gestational age and birthweight.”	High	There are missing data (more than 10% of sample).	High	Cannot locate protocol.
Robledo-Colonia et al. (2012)	Low	“Randomization was performed using a permuted block design with a block si e of 10 and exp:con ratios of 5:5, 6:4 or 4:6.”	High	“Participants and therapists administering the intervention were not blinded.”	Low	“The investigators responsible for outcome assessment were blinded to group allocation.”	Low	“Analysis was according to the principle of intention-to-treat.”	Unc.	The variables don't coincide in the protocol and in the article methodology.
Mosquera-Valderrama et al. (2012)	Low	“Después de la realización de estas pruebas, las pacientes fueron asignadas aleatoriamente.”	High	“La principal limitación es que los terapeutas de campo y las participantes no pueden ser cegados a las intervenciones con el entrenamiento físico aeróbico.”	Unc.	There is no evidence about blinding of outcomes, but everything suggests that it is done.	Low	“…destacamos la validez de los hallazgos debido al diseño del estudio, que incorpora varias características que minimizan la posibilidad de sesgo en los resultados, tales como la aleatorización, y el análisis de intención de tratamiento.”	Unc.	The variables don't coincide in the protocol and in the article methodology.

Unc., unclear.